User location-aware edge services selection based on generative adversarial network and improved ant colony algorithm

Xiuguo Zhang,Shasha Tian,Yufei Liu,Zhiying Cao

DOI: https://doi.org/10.1007/s10489-022-04093-z

IF: 5.3

2022-10-15

Applied Intelligence

Abstract:In a mobile edge environment, a mobile user's location may change at any time. So the service coverages of current edge servers may exceed the mobile user's location, which prevents the user from getting the service they need. For some low latency services, such as video optimized transmission and Internet of vehicles, it is necessary to maintain business continuity by timely selecting services provided by appropriate edge servers during the rapid movement of users. Therefore, this paper proposes a user location-aware edge services selection method based on generative adversarial network (GAN) and improved ant colony algorithm. Using this method to select edge services can provide users with uninterrupted services and improve service quality and user satisfaction. Firstly, a mobile user location prediction method based on the generative adversarial network and the attention mechanism is proposed. This method adopts graph attention network (GAT) and recurrent neural network (RNN) to model user in space and time dimensions, and uses GAN to predict the user's next moment location, according to which, the edge services meeting user's needs can be selected in advance. Then, an edge services selection method based on the predicted user location and improved ant colony algorithm is proposed. This method adopts ant colony algorithm with improved pheromone updating rules and single-cycle optimal list to select edge services, which enhances the global search ability of ants, accelerates the convergence speed of the algorithm and avoids the algorithm falling into a local optimal solution. Experimental results show that the proposed user location prediction method is more reasonable and accurate in predicting user location. The proposed edge service selection method has obvious advantages in aspects of algorithm convergence speed, accuracy and service response time.

computer science, artificial intelligence